@Test
public void testPruningThreshold() {
SimpleDataset dataset;
String exampleString =
"Relevant |BT:tree| (NOTYPE##ROOT(NOTYPE##S(NOTYPE##NP(NOTYPE##NNP(LEX##good::n))(NOTYPE##NNP(LEX##bank::n)))))|ET| ";
dataset = getDatasetFromString(exampleString);
List<List<Double>> info = new ArrayList<List<Double>>();
ArrayList<Double> weights = new ArrayList<Double>();
weights.add(0.1);
weights.add(0.4);
info.add(weights);
TreeAddAdditionalInfoFromArray weigthLoader =
new TreeAddAdditionalInfoFromArray(new TreeNodeSelectorAllLeaves(), info, "weight", "tree");
PruneNodeLowerThanThreshold nodePrunerChecker =
new PruneNodeLowerThanThreshold(0.3, "weight", true, 1.0);
TreeNodePruner treePruner = new TreeNodePruner(nodePrunerChecker, "tree");
System.out.println("Removing all nodes whose value of the field -weight- is lower than 0.3");
System.out.println(weigthLoader.describe());
System.out.println(treePruner.describe());
dataset.manipulate(weigthLoader);
dataset.manipulate(treePruner);
String treepairAfter =
"Relevant |BT:tree| (NOTYPE##ROOT(NOTYPE##S(NOTYPE##NP(NOTYPE##NNP)(NOTYPE##NNP(LEX##bank::n)))))|ET| ";
System.out.println("Pruned sentence:" + getStringFromDataset(dataset));
System.out.println("Expected result:" + treepairAfter);
Assert.assertEquals(getStringFromDataset(dataset), treepairAfter);
}
public static void pruningMaxSentenceLength(
SimpleDataset dataset,
int maxNumberOfLeavesPerSubtree,
SelectRepresentationFromExample treeSelector) {
PruneNodeLeafNumber sentenceLengthPruner = new PruneNodeLeafNumber(maxNumberOfLeavesPerSubtree);
PruneNodeIfLeaf internalNodePruner = new PruneNodeIfLeaf();
TreeNodeSelectorAllChildren sentenceRoots = new TreeNodeSelectorAllChildren();
TreeNodePruner sentencePrunerClass =
new TreeNodePruner(
sentenceLengthPruner,
treeSelector,
internalNodePruner,
sentenceRoots,
TreeNodePruner.UNLIMITED_RECURSION);
System.out.println(sentencePrunerClass.describe());
dataset.manipulate(sentencePrunerClass);
}
@Test
public void testMaxNumberOfSentencesPruning() {
SimpleDataset dataset;
String exampleString =
"Relevant |BT:tree| (NOTYPE##ROOT(NOTYPE##S(NOTYPE##NP(NOTYPE##NNP(LEX##good::n))(NOTYPE##NNP(LEX##bank::n))))(NOTYPE##S(NOTYPE##NP(NOTYPE##WDT(LEX##which::w)))(NOTYPE##VP(NOTYPE##VBZ(LEX##be::v)))(NOTYPE##NP(NOTYPE##DT(LEX##a::d))(NOTYPE##JJ(LEX##good::j))(NOTYPE##NN(LEX##bank::n)))(NOTYPE##PP(NOTYPE##IN(LEX##as::i)))(NOTYPE##PP(NOTYPE##IN(LEX##per::i)))(NOTYPE##NP(NOTYPE##PRP$(LEX##your::p))(NOTYPE##NN(LEX##experience::n)))(NOTYPE##PP(NOTYPE##IN(LEX##in::i)))(NOTYPE##NP(NOTYPE##NNP(LEX##doha::n)))))|ET|";
dataset = getDatasetFromString(exampleString);
PruneNodeNumberOfChildren childrenPruner = new PruneNodeNumberOfChildren(1);
TreeNodePruner treePruner = new TreeNodePruner(childrenPruner, "tree", null, 0);
System.out.println("Removing all sentences but the first one");
System.out.println(treePruner.describe());
dataset.manipulate(treePruner);
String treepairAfter =
"Relevant |BT:tree| (NOTYPE##ROOT(NOTYPE##S(NOTYPE##NP(NOTYPE##NNP(LEX##good::n))(NOTYPE##NNP(LEX##bank::n)))))|ET| ";
System.out.println("Pruned sentence:" + getStringFromDataset(dataset));
System.out.println("Expected result:" + treepairAfter);
Assert.assertEquals(getStringFromDataset(dataset), treepairAfter);
}
